Cluster file storage systems often use a distributed lock manager (DLM) for accessing shared data stored on one or more disks. Different types of read and write locks have thus been designed for DLMs to implement these locks with respect to a variety of data, including that which is stored in database files. Therefore, to access a disk block for example, a lock may be issued by the DLM, so that the disk block can be accessed while the lock is in place. This can help provide access to valid data, while avoiding race conditions.
When multiple nodes attempt to access the same data, read and write lock access requests are usually queued. This can lead to a deadlock situation. As a result, the use of conventional DLM operations often includes taking to use read locks, and to resolve deadlocks.